Author(s): Deepak Krishnamurthy

Originally published on Towards AI.

A classifier approach to Carnatic Raga identification

One of the main sub-genres of Indian Classical music is Carnatic music. Predominantly sung in the south of India, it is one of the oldest and most complex musical systems consisting of Śruti (the foundational pitch or drone), Svara (the individual musical notes), Rāga (the complete melodic framework), and Tāla (the rhythmic framework). Of these, the Rāga or Raga is the melodic framework with its own rules and signature. For a trained carnatic connoisseur, listening to a song and identifying the raga (or framework/backbone that supports that song) is a piece of cake that comes with years of practice to identify and decipher the subtle microtones (gamakas) that distinguish one raga from another, especially if the ragas are closely associated with each other.

This article discusses the development and implementation of a machine learning model to identify Carnatic ragas using an existing music genre identification model called VGGish. The author explores the creation of a dataset from YouTube instrumental renditions, focusing on three specific ragas for analysis. By leveraging pre-trained deep learning techniques, the model achieves significant classification accuracy, though it struggles with identifying similarities between closely related ragas, particularly Shankarabharanam and Kalyani. The findings highlight the model’s potential for extending raga identification to various forms of Carnatic music, while also addressing challenges in classification precision.

Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI